Robust 3D pose estimation of artificial knee implants from X-ray fluoroscopic images

To achieve 3D kinematic analysis of artificial knee implants, 2D/3D registration technique, which uses X-ray fluoroscopic images and computer-aided design (CAD) model of the knee implant, have been applied to clinical cases In previous study, spurious edges and noises from the edge detection of the implant silhouette were erased manually, and that was one of the problems for practical clinical applications. This study presents a robust method called iterative weighted least square method, which estimates 3D pose of the implants in a stable, even when the spurious edges and noises exist. In order to ensure the validity of stable 3D pose estimation of the implants, experiments using actual fluoroscopic images including noises were performed. Results of experiments showed that the estimation accuracy and stability was sufficient in the images that include noises normaly, and the feasibility and effectiveness of present method was demonstrated.